EP3815034A1 - Détection de filigranes d'image semi-transparents - Google Patents

Détection de filigranes d'image semi-transparents

Info

Publication number
EP3815034A1
EP3815034A1 EP19773283.7A EP19773283A EP3815034A1 EP 3815034 A1 EP3815034 A1 EP 3815034A1 EP 19773283 A EP19773283 A EP 19773283A EP 3815034 A1 EP3815034 A1 EP 3815034A1
Authority
EP
European Patent Office
Prior art keywords
image
encoded
possibly
pixels
watermarking
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
EP19773283.7A
Other languages
German (de)
English (en)
Inventor
Dake He
Qichen PAN
Kui PAN
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Google LLC
Original Assignee
Google LLC
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Google LLC filed Critical Google LLC
Publication of EP3815034A1 publication Critical patent/EP3815034A1/fr
Pending legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/46—Embedding additional information in the video signal during the compression process
    • H04N19/467—Embedding additional information in the video signal during the compression process characterised by the embedded information being invisible, e.g. watermarking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T1/00—General purpose image data processing
    • G06T1/0021—Image watermarking
    • G06T1/005—Robust watermarking, e.g. average attack or collusion attack resistant
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T1/00—General purpose image data processing
    • G06T1/0021—Image watermarking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/44—Decoders specially adapted therefor, e.g. video decoders which are asymmetric with respect to the encoder
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2201/00—General purpose image data processing
    • G06T2201/005—Image watermarking
    • G06T2201/0051—Embedding of the watermark in the spatial domain
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2201/00—General purpose image data processing
    • G06T2201/005—Image watermarking
    • G06T2201/0061—Embedding of the watermark in each block of the image, e.g. segmented watermarking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2201/00—General purpose image data processing
    • G06T2201/005—Image watermarking
    • G06T2201/0065—Extraction of an embedded watermark; Reliable detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2201/00—General purpose image data processing
    • G06T2201/005—Image watermarking
    • G06T2201/0083—Image watermarking whereby only watermarked image required at decoder, e.g. source-based, blind, oblivious
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/10—Recognition assisted with metadata

Definitions

  • first-party content providers can provide information for presentation in electronic documents, for example web pages : or application interlaces.
  • the documents can include first-party content provided by first-party content providers and third-party content provided by third-party content providers.
  • Third-party content can be added to an electronic document using various techniques. Some documents include tags that instruct a client computer at. which the document is presented to request third-party content items directly from third-party content providers. Other documents include tags that instruct the client computer to call an intermediary service that partners with multiple third-party content providers to return third- party content items selected from one or more of the third-party content pro viders. In some instances, third-party content items are dynamically selected for presentation in electronic documents, and the particular third-party content items selected for a given serving of a document may differ from third-party content items selected for other servings of the same document.
  • This specification describes computer-based systems, methods, devices, and other techniques for predicting whether ah image likely is or is not encoded with a visually imperceptible watermark image.
  • substantial computational time and expense may he saved in a decoding process by avoiding attempts to decode images for which no watermark has been detected.
  • a service system is called upon to identify metadata associated with a source image.
  • the source image may be an image that was served by a third-party content provider and presented in a webpage or in a graphical interface of a native application at a user computing device (although source images may originate in various other contexts as well).
  • the source image may be rendered in a context where a watermarking image is usually encoded in the source image through alpha-blending of the watermarking image over the source image.
  • a QR-code may be converted to a watermarking image by interspersing blank, non-encoded pixels so that each original (encoded) pixel In the QR-code directly neighbors all blank pixels.
  • the QR-code may include or reference information: about the origin of the source image, the time at which it was served * and/or other pertinent information or metadata in some cases, screenshots of the source images may he captured and provided to the sendee system for decoding, e,g., to recover the QR-code and its encoded data,
  • screenshots may depict source images that have not been encoded with a watermark image.
  • Other screenshots may depict images that have been encoded with a watermark image, but for: which the encoding has been corrupted (e.giller due to the presence of noise from image compression or due to intentional obfuscation/attacks) to such a degree that the prospect of successful decoding becomes impossible or prohibitively expensive.
  • the nature of the encoding and corresponding decoding process itself which involves estimation of a pixel's value in the watermarking image based on the shade or coloration of a corresponding pixel in the screenshot image relative to neighboring pixels, can be imprecise and render some encoded source images incapable of being decoded even in the absence of noise.
  • the techniques described in this specification provide a predictive model for analy zing a unique set of features of the source image that allow for generation of a prediction indicative of whether the source image likely does or does not con tain a watermarking image.
  • the prediction can then be employed for a number of purposes, including determining whether to attempt decoding of the source image and/or determining whether an unsuccessful decoding was likely the result of the source image not containing a watermark that could be decoded in the first instance, or was likely the result of noise corrupting the watermark or another reason.
  • the methods can include obtaining, by a computing system, a first image, such as a screenshot of a source image that was captured at a user (e.g Stephen client) computing device or a scaled version of the screenshot/souree image.
  • the computing system analyzes attributes (e.g Thompson colors, shade) of pixels in the fust image to determine encoded values of possibly encoded pixels in the first image.
  • the system analyzes patterns of encoded values of the possibly encoded pixels in the first image to determine possibly encoded regions of the first image.
  • the system can then determine values for features of the possibly encoded regions of the first image
  • the values for the features of the possibly encoded regions of the first image are provided to and processed with a predictive model to generate a prediction indicative of whether a watermarking image is encoded in the first image.
  • the system can then apply and output the prediction in one or more ways.
  • Generating the prediction can include selecting a binary value that indicates either that the watermarking image is encoded in the first image or the watermarking image is not encoded in the first image.
  • Generating the prediction can include generating a numerical prediction that quantifies a likelihood that five watermarking image is or is not encoded in the first image.
  • the prediction can include at least one of storing the prediction in a memory of a computing system or providing the prediction for presentation to a user,
  • Outputting the prediction can include providing the prediction to a controller that is configured to determine, based at least on the prediction, whether to apply a decoding process to the first image or a second image from which the first image was derived to recover fire watermarking image.
  • the controller can be configured to apply the decoding process to the first image or the second image to recover the watermarking image in response to identifying that the prediction indicates the watermarking image is encoded in the first image
  • the controller can be configured to not apply the decoding process to the first image or the second image to recover the watermarking image in response to identifying that the prediction indicates the watermarking image is not encoded in the first image,
  • Obtaining the first image can include obtaining a screenshot image from a user's computing device, the screenshot image showing sewed content in a webpage or a graphical user interface of a native application,
  • the methods can further include actions of obtaining a second image that was rendered at an unknown zoom level; identifying a plurality of possible zoom levels for the second image; and generating a plurality of scal ed images incl uding the first image from the second : image, wherein generating each scaled image includes (i) selecting a different possible zoom le vel from the plurality of zoom levels, and (if) sealing the second image by an amount that would be required to realize a common target zoom level if the second image had been rendered at the different possible zoom level.
  • the methods can further include, far each scaled image in the plurality of scaled images; analyzing colors of pixels in the sealed image to determine encoded values of possibly encoded pixels in the scaled image; analyzing patterns of encoded values of the possibly encoded pixels in the sealed image to determine possibly encoded regions of the scaled image; determiningvalues for fea tures of the possibly encoded regions of the scaled image; and determining to provide the values for features of the possibly encoded regions of the first image .for processing by the predictive model rather than the values for features of the possibly encoded regions of other ones of the plurality of scaled images based on one or more criteria.
  • Analyzing colors of pixels in the first image to determine the encoded values of the possibly encoded pixels in the first image can include, for each of all or some of the possibly encoded pixels in the first image, comparing a color of the possibly encoded pixel to the respecti ve colors of one or more pixels adjacent to the possibly encoded pixel in the first image.
  • Analyzing the patterns of encoded values of the possibly encoded pixels in the first image to determine the possibly encoded regions of the first image can inc lude identify ing regions that contain one or more blocks of possibly encoded pixels whose encoded values are determined to possibly correspond to one or more encoded instances of the watermarking image in the first image.
  • Identifying regions of the first image that contain one or more blocks of possibly encoded pixels whose encoded values are determined to possibly correspond to one or more encoded instances of the watermarking image in the first image can include locating instances of a pre-defined bit pattern that occur in each block.
  • Determining the values for the features of the possibly encoded regions of the first image can include; for each possibly encoded region, counting a number of blocks of possibly encoded pixels in the possibly encoded region whose encoded values are determined to correspond to a different encoded instance of foe watermarking image; wherein the features of the possibly encoded regions of the first image are selected from a : group comprising a mean number of coun ted blocks among the possibly encoded regions, a maximum number of counted blocks among the possibly encoded regions, a number of po ssib ly encoded regions i denti fied, and a s tandard deviation in foe number of counted blocks among the possibly encoded regions.
  • the first image can be a scaled version of a second image.
  • the methods can further include actions of: mapping the possibly encoded regions of the first image to corresponding regions of the second image; determining variances in each of one or more color channels among the pixels in each mapped region of the second image; for each color channel, determining the minimum variance for the color channel among the mapped regions of the second image; and generat ing the predicti on in d icative of whether a watermarking image is encoded in the first image comprises further processing the minimum variances for each of the one or more color channels among the pixels in each mapped region of the second image.
  • the predictive model can include a binary classification model.
  • the predictive model can include a neural network
  • Some implementations of the subject matter disclosed herein include corresponding computer systems, apparatus, and computer programs recorded on one or more computer-storage devices (e.g., computer-readable media), each configured to perform the actions of the methods.
  • a system of one or more computers in one or more locations can he configured to perform particular operations of actions by virtue of having software, .firmware,hardware, or a combination of them installed on the system that in operation causes or cause the system to perform the actions.
  • One or more computer programs can be configured to perform particular operations or actions fiy virtue of including instructions that, when executed by data processing apparatus (e.g., one or more processors), cause the apparatus to perform the actions.
  • a system may generate a watermark image to augment an arbitrary source image, even if the source image is not known when the watermark image is created. Because the watermark image may be overlaid on the source image for presentation to a user at a client computer remote from the system, the system need not directly obtain or modify the source image before the watermark image is transmitted to a client computer.
  • the system may recover a watermark image, or may recover an encoding image from which the watermark image was derived, even without a priori knowledge of the source image, thereby enabling a system fo decode arbitrary images without knowing where they originated from.
  • the watermark image can be substantially transparent so as to he imperceptible to users when the watermark image is rendered over a source image so that the watermark image does not degrade the percei ved visual quality of the source image.
  • a predictive model can be employed to predict whether a possibly encoded source image likely has been or has not been encoded with a watermarking image. By first generating such a prediction before atempting to complete a decoding process on the source image, computational time and expense may be mitigated by avoiding atempts to decode images that likely were never encoded in the first instance.
  • Figure 1 is a block diagram of a computing environment for augmenting electronic documents served to client devices with watermark images arranged for display over one or more source images in the documents and a server system configured to implement operations for watermark prediction and decoding.
  • Figures 2A-2J illustrate example techniques for generating a semi-transparent watermark image and blindly encoding the watermark image over a source image.
  • Figure 3 is a flowchart of au example process for augmenting source images presented in documents served to a client computer with a watermark image, and subsequently decoding a screenshot containing a source image that has been encoded with the watermark.
  • Figure 4 is a flowchart of an example process for generating a watermark image from an encoding image.
  • Figure 5 is a depiction of an example encoding image in the form of a two- dimensional binary quick-response code (QR-code).
  • QR-code binary quick-response code
  • Figure 6 is a flowchart of an example process for predicting whether a captured source image (e.g , a screenshot) is encoded with one or more watermarking images.
  • a captured source image e.g , a screenshot
  • Figure 7 is a flowchart: of an example process for identifying regions in an image that have possibly been encoded with one or more watermarking images.
  • Figure 8 is a diagram of a watermark predictive model that processes values of various features associated with a possibly encoded source image to generate a watermark prediction.
  • Figure 9 provides a depiction of unsealed and scaled versions of a possibly encoded source image along with indications of possibly encoded regions identified in the possibly encoded source image.
  • Figure 10 depicts t he schema of an example database table for st oring a possibly encoded source image for a set of zoom levels.
  • Figure 11 is a flowchart of an example process for training a predictive model to generate predictions indicati ve of whether a possibly encoded source image likely is or is not encoded with a watermarking image
  • Figure 12 is a block diagram of a computing system that can be used in connection with computer-implemented methods and other techniques described in this document,
  • This specification describes techniques for detecting semi-transparent watermark images that have been blended over a source imageso as to render the watermark visually imperceptible to a human user under normal viewing condi tions.
  • the detection process can employ a predictive model that; processes features of a possibly encoded source image to generate a prediction indicating whether the image likely does or does not contain watermarks.
  • adiagram of a computing environment 100 is depi cted in which source images are watermarked to digitally encode information associated with the source image, and in which encoded source images are decoded to recover the watermark or to access information referenced by the watermark.
  • the environment 100 includes a server system 102; a client computer 104, and computing systems for one or more source image providers 106a-n.
  • the server system 102, client computer 104, and source image providers 106a-n are cormmmicahly connected over one or more networks such as the Internet or a local area network (LAN) in general, the client computer 104 is configured to generate and transmit requests for electronic documents to the server system 102, Based on the requests from the client computer 104, the server system 102 generates responses (e.g,, electronic documents) to return to the client computer 104.
  • a given response can reference a selected source image 128a that is to be rendered at the client computer 104 for display to a user, where the source image 128a is provided by one of the source image providers 106a-n.
  • the server system 102 can augment the response served to the client computer 104 with a distinct, semi-transparent watermarking image 126 that is arranged for display in a presentation of the response document at the client computer 104 over the source image 128.
  • the server system 102 may he configured to employ a source image encoder 138 to watermark or encode the source image 128 with the watermarking image 126 at the server system 102 in a same or similar manner to the encoding that results from client computer 104 lay ing the watermarking image 126 over source image 128 (e.g., by alpha-blending the semi-transparent watermark image 126 in the foreground over source image 128), thereby merging the source image 128 and watermarking image 126 into a single encoded source image that is served to the client computer 104.
  • the foll owing discussion of Figure 1 focuses on the former implementations where a distinct watermark image 126 is served to the client computer 104 for rendering by the client computer 104 over top of
  • the client computer 104 can generally be any type of computing de vice that presents images and other content to one or more human users.
  • the client computer 104 may include an application, such as a web browser application, that makes requests to and receives responses from the server System 102, The application may execute a response from the server system 102, such as web page code or oilier types of document files, to present the response to the one or more users of the client computer 104,
  • the client computer 104 includes an electronic display device (e.g., an LCD or LED screen, a CRT monitor, a head-mounted virtual reality display, a head-mounted mixed-reality display), or is coupled to an electronic display device, that displays content from the rendered response to the one or more users of the client computer 104.
  • the displayed content can include the selected source image 128a and the watermarking image 126 displayed over top of the selected source image 128a in a substantially transparent manner.
  • the client computer 104 Is a notebook computer, a smartphone, a tablet computer, a desktop computer, or a smartwatch or other wearable device.
  • the source image 128 provided in the response to the client computer 104 is a third-party content item that, for example, is not among content provided by a first-party content provider of the response.
  • the creator of the web page may include a slot that is configured to be populated by a source image from a third-party content provider that differs from the creator of the web page (e g., a provider of an image repository).
  • the first-party content provider may directly link to a third-party source image 128,
  • the client computer 104 may request the source image 128 directly from acorresponding compu ting system for one of the source image providers 106a-n or indirectly via an intermediary service, such as a service provided by server system 102 or another server system.
  • the server system 102 is configured to respond to a request from the client computer 104 with an electronic document and a semi-transparent watermarking image .126 that is to be displayed hr the electronic document over a source image 128a.
  • the server system 102 and each of its components can be implemented as one or more computers, including data processing apparatus, in one or more locations.
  • the server system 102 can include one or more of a front-end subsystem 108, an image generation subsystem 110, a response formatter 116, an image analysis and decoder module 118, a response records database 120, a watermark predictive model 132, a model training engine 134, a controller 136, and a source image encoder 138.
  • Bach subsystem can be implemented on hardware including data processing apparatus such as one or more processing devices.
  • Each of the components 108-120, 132-138 is configured to perform the respective operations described herein.
  • each component 108-120 and 132-138 may be defined by instructions stored in memory on one or more compute-r-readable storage devices, and the operations may be earned out when the instructions are executed fey one or more processors of the server system 102.
  • foe operations may be implemented using software, hardware, firmware, or a combination of software, hardware, and firmware.
  • some operations are described herein as being performed by a specific one of the components 108-120 or 132-138 by way of example, in other implemen tations, some or all of the operations of two or more of the components 108-1:20 may be consolidated and performed instead by a single component.
  • the operations of any one of the components 108- 120 or 132-38 may fee divided among two or more components.
  • server system 102 can be configured to carry out the techniques illustrated in Figures 2A-2 ⁇ and the processes 300, 400, 600, 700, and 1100, The operations of processes 300, 400, 600, 700, and 1100, and other techniques, are described in further detail below with respect to Figures 3, 4, 6, 7, and 11.
  • An overview of foe server system 102 aed its components 108-120, 132-138 follows below, with further detail of the operations and other techniques perforated by these components described subsequently with respect to Figures 2-1.1.
  • the ftont-end subsystem 108 provides an interface for communicating over one or more networks.
  • the front-end subsystem 108 recei ves requests from the client computer 104 and transmits responses to the requests, along with any content associated with foe requests such as a watermarking image and, optionally, a source image, to the client computer 104.
  • the front-end subsystem 108 may also communicate with the computing systems of source image pro viders 106a-n, e.g., to obtain a source image 128a to serve to the client computer 104.
  • the front-end subsystem 108 may also communicate with, and include a controller 136 far coordinating activities among, each of the components 112-120, 132-138 of the server sy stem 102.
  • the front-end subsystem 108 can include a wired (e.g,, metallic or optical), wireless, or combination of wired and wireless communications interlaces that enable the front-end subsystem to connect to a suitable communications network.
  • the image generation subsystem 110 is configured to generate images from input data, in particular, the image generation subsystem Tit) includes an encoding input generator 112 and a watermarking image generator 114.
  • the encoding input generator 112 processes a pla e data item 122 to generate an encoding image 124 that encodes the plaintext data item 12:2
  • the plaintext data item 122 can be any data that is capable of being encoded within the constraints of the encoding input generator 112.
  • the plaintext data item 122 may be a text sample: or string with a maximum length ofn characters, since the size of the encoding image 124 may be capable of providing lossless encoding for text samples only up to the pre-deilned maximum length of n characters.
  • the plaintext data item 122 includes a session identifier that uniquely identifies a network session between the client computer 104 and the server system 102 during which a response is served to a request from the client computer 104,
  • the plaintext data item 122 includes or references source image data that identifies the particular source image 128a served to the client computer 104 or information associated with the source image 128a (e.g., information that indicates which of the source image providers 106a-n provided the particular source image 128a served to the client computer 104 and a timestamp indicating when the source image 128a was served or requested).
  • the response records database 120 stores data that correlates such information about a source image 128 or a response sewed for a particular request, in order to make the detailed information accessible via the session identifier or other information represented by the plaintext data item 122.
  • the response records database 120 can also associate a session identifier with source image data, thereby making the source image data accessible by querying the database 120 using the session identifier represented by the plaintext data item 122 A user can then identify, for example, which of the source images 128a-n was served to the client computer 104 at what time and from which source image provider 106a-n for using the session identifier from the plaintext data item 122,
  • the encoding image 124 is an Image that encodes the plaintext data item 122.
  • the encoding image 124 is a matrix-type barcode that represents the plain text data item: 122.
  • a suitable matrix-type barcode is a Quick Response Code (QR code).
  • QR code Quick Response Code
  • the encoding image 124 can have a pre-defined size in terms of a number of rows and columns of pixels.
  • Each pixel in the encoding image 124 can encode a binary bit of data, where the value of each bit is represented by a different color. For example, a pixel that encodes the binary value "1" may be black while a pixel that encodes the binary value "0" may be white.
  • the smallest encoding unit of an encoding image 124 may actually be larger than a single pixel.
  • the smallest encoding unit is assumed to be a single pixel it should be appreciated, however, that the techniques: described herein may be extended to implementations where the smallest encoding unit is a set of multiple pixels, e.g,, a 2x2 or 3x3 set of pixels.
  • the image generator subsystem 110 further includes a watermarking image generator 114,
  • the watermarking image generator 114 is configured to process the encoding image 124 to generate a semi-transparent w atermarking image 12b.
  • the semi-transparent watermarking image 120 is derived from the encoding image 124 and also encodes the plaintext data item 122.
  • the transparencies, colors, arrangement of encoded pixels and/or other features of the watermarking image 126 may be changed from the transparencies, colors, arrangement of encoded pixels and/or other features of the encoding image 124.
  • the encoding image 124 may be uniformly opaque and consist of encoded pixels that are closely packed adjacent to each other
  • the watermarking image 126 may include some fully transparent pixels and some partially transparent pixels.
  • the encoded pixels in the watermarking image 126 may be spaced relative to each other so that each encoded pixel is surrounded by non-encoded pixels (i.e,, “blank” pixels).
  • the transformation of the encoding image 124 to the watermarking image 1:26 may be performed so that, after the watermarking image 126 is overlaid and merged on a background source image 128a, the encoded information may he decoded, e.g,, by reconstructing the encoding image 124 or the watermarking image 126.
  • the watermarking image 126 may be generated directly from the plain text data 122 without explicitly generating the encoding image 1:24 as an intermediate operation on the way to achieving watermark image 126.
  • the server system 102 employs source image encoder 138 to directly merge the watermarking image 126 over top of the source image 128a for service of the merged image to the client computer 104, the server system 102 may directly encode the watermark in the source image 128 without explicitly generating the encoding image 124, watermarking image 126, or both.
  • the response formatter 116 is configured to generate a response to return to the client computer 104 in reply to the client's request for an electronic document.
  • the response can include one or more content items, including first-party content items and third party content items, which collectively form an electronic document such as a web page, an application interface, a PDF, a presentation slide deck, or a spreadsheet.
  • the response includes a primary document that specifies how various content items are lo be arranged and displayed.
  • the primary document such as a hypertext markup language (HTML) page
  • the primary document may refer to first-party content items and third-party content items that are to he displayed in the presentation of the document in some implementations.
  • the response formatter 116 is configured to add computer code to the primary document that instructs the client computer 104, when executing the response, to display one or more instances of the watermarking image 126 over the source image 128a, e.giller to add a watermark to the source image 128a that is substantially imperceptible to human users.
  • the application at the client computer 104 that renders the electronic document can perform a blending technique to overlay the watermarking image 126 on the source image 328a according to the specified transparencies oftbe watermarking image 126
  • the response formatter 116 may add code that directs the client computer 104 to display the source image 128a as a background image in a third-party content slot in an electronic document and to display one or more instances of the watermarking image ! 26 as a foreground image over the source image 128a.
  • the server system 102 further includes an image analysis and decoder module 118.
  • the image analysts and decoder module 118 is configured to decode an encoded/watermarked source image, e.g., to reconstruct encoding image 124 or watermarking image 126, to recover the non-encoded versi on of source image 128a, and/or to reco v er plaintext data 122 so that information from the plaintext data 122 can be read and/or an appropriate record from response records database 120 can be accessed based on the information from plaintext data 122.
  • the encoded/walermarked source image is an image that results from the client computer 104 rendering the watermarking image 126 over the source image 128a.
  • the encoded source image 130 processed by the image analysis and decoder module 118 may be a merged image sho wing the watermarking image 126 blended over the source image 128a.
  • a user of the client computer 104 may receive an inappropriate or irrelevant source i mage 128a from one of the source image providers 106a-n in response to a request for an electronic document.
  • the user may capture a screenshot of the encoded source image 130 and transmit the screenshot to the server system 102 for analysis, e.g., to inquire about the origin of the source image 128a Because the screenshot shows the original source image 128a overlaid by the watermarking image 126, the image analysis and decoder module 118 can process the screenshot to recover an encoded representation of the plaintext data item 122, which in turn can he decoded to recover the plaintext data item 122 itself.
  • the system 102 can then use the recovered plaintext data item 122 for various purposes, e.g., to query the response records database 120 to lookup detailed information about the source image 128a and its origins, or other information about the particular clientsession in which the source image 128a was served to the client computer 104
  • the encoded representation of the plaintext data item 122 that the image analysis and decoder module 118 generates can be, for example, a recovered watermarking image 126 or a recovered encoding image 124.
  • the server system 102 may receive requests to analyze possibly encoded/watermarked source images
  • the term “possibly” refers to: a condition of an item that might be attributable to the item but that is nonetheless unknown to a processing entity (e.g.. server system 102 / decoder 118) that processes the item. That is, the possible condition of an item is a candidate condition of an item for which its truth is unknown to the processing entity.
  • the processing entity may perform processing to identify possible (candidate) conditions of an item, to make a prediction as to the truth of a possible (candidate) condition, and/or to identify possible (candidate) items that exhibit a particular condition.
  • a possibly encoded source image is a source image that is possibly encoded with a watermark but for which it is initially unknown to the server system 102 whether the image actually has been watermarked.
  • the possible encod ing of the source image with a watermark is thus a candidate condition of the source image, and the source image is a candidate item exhibiting the condition of being encoded with a watermark.
  • the possibly encoded source image may result from a user capturing a screenshot of the sotiree image anti providing the image to server system 102 for analysis, but without more information that would indicate a confirmation as to whether the image had been encoded/watermarked.
  • the decoder 1 IS may cooperate with the watermark predictive: model 13:2 and the controller 136 to predict whether the possibly encoded source: image likely does or does not contain a watermark before the decoder 118 completes one or more attempts to decode the possibly encoded source image.
  • the decoder 118 may identify possibly encoded regions of the possib ly encoded source image, and may determine : values for Features of the possibly encoded source image.
  • the values for these features may then be processed by the watermark predictive model 132 to generate a prediction indicative of whether the possibly encoded source image has been encoded with a- watermarking image 126
  • the controller 136 may use this prediction to determine whether decoder 118 should attempt to complete a decoding process on the possibly encoded source image.
  • the watermark predictive model 132 can be implemented on one or more computers in one or more locations where the server system 102 is implemented.
  • model 132 is a binary classifier.
  • model 132 is a neural network (e.. a feedforward neural network having an input layer of operations, one or more hidden layers, and an output layer) or other machine-learning model.
  • the model 132 may be trained on sample training data by model training engine 134.
  • Figures 2A-2I illustrate example techniques for generating a watermarking image encoding plaintext data, and for decoding an encoded source image to recover an encoded representation of the plaintext data ⁇ e.g., to reconstruct one or more of the encoding image, watermarking image, or plaintext data from the encoded source image).
  • Figures 2A-2I is discussed in forther detail below with respect to the processes represented in the flowcharts of Figures 3 and 4.
  • Figure 3 is a flowchart of an example process 300 for using a semi-transparent watermarking image to augment a source image presented in an electronic document served to a client computer.
  • the process 3011 can be carried o ut by a system of one or more computers, e.g.
  • the process 300 begins at stage 302, where thesystem receives plaintext data, e.g , plaintext data item 122.
  • Plaintext data is generally any data that is capable of being encoded in an encoding image such as a QR code.
  • the plaintext data is a session identifier that uniquely identifies a logical session between a client computer and the system, or that identifies an electronic document served in response to a request from a client computer in some implementations, the plaintext data includes or references source image data that identifies a particular source image served to the client computer or information associated with the source Image.
  • an encoding input generator e.g, encoding input generator 11.2 of the system generates an encoding image from the plaintext data.
  • the encoding image Is an image that encodes the plaintext data.
  • the encoding image is a matrix-type barcode that represents the plaintext data.
  • a suitable matrix “ type barcode is a Quick Response Code (QR-cede) such as the QR-code 500 depicted in Figure 5.
  • QR-cede Quick Response Code
  • the encoding image can have a pie-defined size in terms of a number of rows and columns of pixels.
  • Each pixel in the encoding image can encode a binary bit of data, where the value of each bit is represented by a different color. For instance, a pixel that encodes the binary value "1" may be black while a pixel that encodes the binary value "O" may be white
  • the encoding image 202 in this example is a two-dimensional barcode that has three rows and three columns of pixels. Of course, other sizes of encoding images are also possible with more or fewer rows and columns of pixels.
  • some pixels have a first binary color, i.e., black, and other pixels have a second binary color, I.e , white.
  • the color of each pixel is defined by a color value.
  • the pixels located at (row, column) coordinates (1, 2) (3, 1 ), and (3, 2), for example, are white, while the other pixels in encoding image 202 are black
  • each pixel may have a transparency value (also referred to as an “alpha” value) that indicates a transparency level of the pixel.
  • the transparency value may be normalized to the range [0, 1] where a value of "O” represents fully transparent, a value of "1” represents fully non-transparent (opaque), and intermediate values between "O” and "1” represent partial transparencies.
  • the encoding image 202 is fully non-transparent such t hat each of the pixels of the encoding image 202 has a tra nsparency value of "1"
  • Image rendering applications can then use transparency values to determine composite colors for pixels that result when a foreground image is overlaid and blended on a background image.
  • the composite pixel that is ultimately rendered for display to a user will have a color value of fgROB * fgA + bgRGB * ( 1 - fgA) .
  • Figure 2F shows an example background image 212 (e.g., a source image), where each pixel has a color value identified by the symbol CnRGB.
  • Figure 2G shows an encoded source image 214 that has been generated using the blending technique described in the preceding paragraph to overlay watermarking image 210 (foreground image) on source image 212 (background image), where the colors of certain encoded pixels in the encoded source image 214 are a blend of partially transparent foreground pixels from the watermarking image : 210 and the corresponding background pixel colors from the source image 212.
  • the color v al ues and expressions shown in Figure 2G assume that shaded pixels in the foreground watermarking image are black and thus the foreground RGB value for the pixel is (0, 0, 0).
  • the full expression for the composite pixel that results from blending foe foreground watermarking image and the background source image reduces from fgRGB * fgA + bgRGB * (1 - fgA)’ to simply ‘bgRGB * (1 - fgA)2
  • fgRGB * fgA + bgRGB * (1 - fgA) the composite color value for the pixel is defined by the expression Cn RGB * (1 - Ba)
  • the composite color value for the pixel is Cn RGB * (1 - Ba)
  • a watermarking image generator e g , watermarking image generator 114, of foe system generates a watermarking image.
  • the watermarking image is a semi-transparent image that includes pixels having two or more different levels of transparency to encode the plaintext data item.
  • the watermarking image may consist of a first set of pixels having a first transparency le vel, e.g., partially transparent, and a second set of pixels having a second transparency level, e.g, fully transparent.
  • the first set of pixels may be encoded pixels that each represents a respective pixel from the encoding image having a first color value such as black.
  • the second set of pixels may include two subsets of pixels, namely encoded pixels and blank pixels. Although the encoded and blank pixels in the second subset may share the same transparency level, only foe encoded subset of pixels represents pixels from the encoding image having a second color value, such as white.
  • the blank pixels do not represent any pixels from the encoding image, bat are instead interspersed among the encoded pixels of the first and second sets of pixels to facilitate recovery of information from an encoded source image that may later result from the watermarking image being overlaid on a background image
  • Tire process 400 is a process for generating a watermarking image from an encoding image.
  • the process 400 can be performed by a system of one: or more computers e.g., server system 102, and specifically by a watermarking image generator, e.g., watermarking image generator 114.
  • the watermarking image generator identifies an encoding image such as encoding Image 202.
  • the watermarking image generator normalizes the encoding image. Normalizing the encoding image can involve mapping the respective color values and transparency values of all or some of the pixels in the encoding image to pre- defined color aad/or transparency values. For example, if in the original encoding image, binary values encoded in the image were distinguished by pixel color, then in the normalized encoding image, binary values encoded in the normalized image may he distinguished by pixel transparency.
  • Figures 2B and 2C depict one example of normalizing an encoding image in particular.
  • Figure 2B shows an example representation 204 of the original encoding image 202 inwhich the black pixels all have the color value BROS and have the transparency value BA and the white pixels all have the color value W RGB and have the transparency value WA. in some implementations, WA and BA are identical.
  • Figure 2C shows a normalized encoding image 206 where the colors and transparencies of the pixels have been transformed so that the black and white pixels are assigned the common color B RGB but are differentiated by their transparencies. The black pixels are assigned the transparency value B A , while the white pixels are assigned the transparency value of‘0’, i.e., felly transparent.
  • the watermarking image generator generates an initial watermark canvas in some implementations, the initial watermark canvas is a preliminary image that has a pre-de!med size and provides a starting point for creation of a watermarking image.
  • An example initial watermark canvas is canvas 208, as shown in Figure 2D.
  • the canvas 208 has the same number of rows as the normalized encoding image 206, but is made to have twice the number of columns as the normalized encoding image 206 Additionally, all the pixels in the initial watermark canvas 208 are blank pixels that have the same color and transparency level.
  • Blank pixels are pixels that do not encode information from the encoding image and do not correspond to any pixels in the encoding image in some implementations, blank pixels have the same transparency as the second set of pixels in the norma li zed encoding image 206.
  • the blank pixels in canvas 208 are fully transparent as indicated by the latter value in the respective parentheticals (color value, transparency value) for each pixel
  • the watermarking image generator uses the normalized encoding image to add encoded pixels to the initial watermark canvas to create a final watermarking image.
  • Encoded pixels ate pixels in the watermarking image which, in contrast to blank pixels, do encode information from the encoding image.
  • Each encoded pixel in the watermarking image corresponds to one of the pixels in the normalized encoding image.
  • the final watermarking image is generated by replacing a subset of blank pixels in the initial watermark canvas with encoded pixels from the normalized encoding image. For example, a given blank pixel in the initial watermark canvas may be fully transparent.
  • That pixel is made to be m encoded pixel, then that pixel is assigned the transparency of a corresponding pixel from the normalized encoding image that is to be encoded.
  • the pixel may remain fully transparent if the corresponding pixel from fire normalized encoding image is fully transparent.
  • the transparency of the watermark pixel may he adjusted to a partially transparent state if the corresponding pixel from the normalized encoding image is partially transparent
  • Figure 2E shows one example of a watermarking image 210 that the watermarking image generator generates from the initial watermark canvas 208 based On the normalized encoding image 206, Each of the pixels in the watermarking image 21.0 that have a patterned background represents an encoded pixel. Each of the pi xels in the watermarking image 210 having a non -patterned white background represents a blank pixel. As the watermarking image 210 shows, the encoded pixels are distributed among the blank pixels such that each encoded pixel is neighbored at the top, bottom, left, and/or right by at least two blank pixels.
  • the watermarking image generator distributes encoded pixels in the watermarking image so that each encoded pixel is directly adjacent to at least one, two, three, four, or more blank pixels. This arrangement can allow encoded information to later be recovered from a source image that has been augmented by a watermarking image without a priori knowledge of the source image, as described further below.
  • the watermarking image generator has interspersed the encoded pixels in the watermarking image by starting with the arrangement of pixels in the normalized encoding image, inserting a blank pixel directly to the right of each encoded pixel in odd-numbered rows, aad inserting a blank pixel directly to the left of each encoded pixel in even-numbered rows so as to shift the encoded pixels in even-numbered rows by one pixel relative to the encoded pixels in add-numbered rows.
  • the effect is to stagger the encoded pixels and to surround them by blank pixels, as shown in the final watermarking image 210.
  • the only pixels that are not fully transparent are the encoded pixels that correspond to black pixels in the original encoding image 202.
  • encoded pixels in the watermarking image 210 that correspond to pixels of a first binary color in the original encoding image 202 may be assigned a first transpatency level, whereas a different, second transparency level may be assigned to both blank pixels and encoded pixels that correspond to pixels of a second binary color from the original encoding image 202.
  • the watermarking image generator can use other techniques for mapping encoded pixels from the normalized encoding image 206 to the final watermarking image 210.
  • the watermarking image generator may use a mapping template that correlates positions for encoded pixels in the watermarking image 210 to pixel positions in the normalized encoding image 206. Every pixel in the normalized encoding image 206 can be mapped to a pixel in the watermarking image 210.
  • the system can also use the mapping template to perform a reverse mapping of encoded pixels from an encoded source image in order to reconstruct the watermarking image 210 or the encoding image 202 front the encoded source image.
  • the mapping template can identify any suitable arrangement of encoding pixels (e,g., in any order) in the watermarking image 210, so long as each of the encoding pixels neighbors blank pixels.
  • the system generates an entry in a response records database, e g. , response records database 120.
  • the response records database contains a log that stores information about each response or other electronic document (e.g., a pushed electronic document that is not responsive to a specific request) transmitted to one or more client computers in communication with the server system over a period of time.
  • Each response may be associated with a client session, which may entail one or more request-response interactions between a given client computer and the server system.
  • a given entry in toe response records database includes a session ID, a response ID, and source image data.
  • the session ID uniquely identifies a server-client network session
  • the response ID uniquely identifies a response served to the client computer during a given session
  • the source image data identifies information about a source image transmitted to the client: computer for a particular session or response.
  • the session ID, response ID, or both may form the plaintext data that is encoded in the encoding image and watermarking Image so that the system can later use a watermark overlaid on a source image to recover the session ID, response ID, or both and to lookup the appropriate source image data associated with: the session IP and/or response IP
  • the server system serves an electronic document to toe client computer.
  • the electronic document may include computer code that, when executed by toe client computer, causes the client computer to request and obtain the source image and the watermarking Image
  • the computer code may include instructions for the client computer to overlay the watermarking image on the source image when the electronic document is rendered for presentation to a user.
  • an example process 600 for rendering an encoded source image at a client computer is depicted In Figure 6
  • the process 600 can he carried out by one or more computing devices, e.g., client computer 104.
  • the process 600 is carried out by an application installed on the client computer, such as a web browsing application.
  • the server system ⁇ 02 may blend the watermarking image and the source image to create the encoded source image at the server system 102 before it is served to the client computer 104.
  • toe server sy stem 102 may be called upon to decode an encoded source image.
  • the system receives an encoded source Image, which may have been captured as a screenshot by a user of the client computer 104 and submitted to the server system 102 for decoding and analysis.
  • the sy stem decodes the encoded source image.
  • the decoding process generally involves identifying the encoded values of the encoded pixels in the encoded source image, e.g., to determine whether each encoded pixel corresponds to a black pixel (value 1) in toe encoding source image (e.g , a QR-code) or a white pixel (value 0) in toe encoding source image.
  • the color of toe pixel may be compared to its nei ghboring pixels, and if the color of toe pixel is darker than its neighboring pixels by a certain amount, then it may be considered to encode a black pixel (value 1) from the encoding image if the color of toe pixel is not darker than its neighboring pixels by toe requisite amount, then it may be considered to encode a white pixel (value 0) from toe encoding image.
  • the same encoded pixel from multiple instances of the watermarking image encoded in the source image may be analyzed and toe results statistically averaged.
  • a machine-learning model may be trained to perform a decoding analysis. Nonetheless, the decoding process can be computationally expensive and not always successful. For example, if the pattern of colors in the underlying source image obscures the darkening imposed by a partially transparent, black encoding pixel, then the encoded value may not be successfully recovered.
  • the system may need to iteratively search across the screenshot of the encoded source image to Identify possible locations for the watermarking image, and may be required to do so for many possible zoom levels or resolutions since the system may not be privy to data indication at what zoom level or resolution the screenshot was rendered and captured. Accordingly, the descriptions of Figures 6-10 provide techniques for increasing efficiency of the decoding process using a predicti ve model to generate a prediction of whether a possibly encoded source image is even encoded with a watermarking image at all.
  • the server system uses the decoded information to recover plaintext data.
  • the system may then use the plaintext data, for speechous ends.
  • the plaintext data represents a session ID or a response ID.
  • the server system can query the response records database using the session ID or response ID to lookup source image data and determine, for example, a particular content provider that provided the source image.
  • Figure 5 depicts ah example QR-code 500 that can serve as an encoding image, e.g,, encoding image 124 for purposes of the techniques described in this specification.
  • the QR-code 500 has a fixed size of 21x21 pixels in this example, although QR-codes of other pre-defined sizes would also be suitable.
  • the 21x21 QR-code 500 will be referenced in the discussion of figures 6-10 to illustrate the techniques disclosed herein for predicting whether a possibly encoded source image likely has or has not been encoded with a watermarking image.
  • a di stinctive feature of the QR-code 500 is its three 7x 7 pixel squares 502a-
  • the square patterns 502a-e aid optical-reading de vices in locating the bounds of QR-code 500 and properly orienting an Image of QR-code 500 so that rows and columns of pixels can Ire ascertained and the code 500 can be successfully read.
  • Each square pattern is defined by seven consecutive black pixels (e.g., encoded value 1) in its first and seventh rows, the pattern black-white-white-white-white-black (e.g., encoded values 1 -0-0-0-0-0-1) in the second and sixth rows, and the pattern black- white-black-black- black- white-black (e.g,, encoded values 1 -0-1-1 -1 -0-1) in the third, fourth, and fifth rows,
  • a watermarking image can be formed from the QR-code 500 as described with respect to Figures 1-4, including by assigning a high-partial transparency value to each black pixel in the code 500, applying a full-transparency value to each white pixel in the code 500, inserting a blank (non-encoded) folly transparent pixel to the right of each pixel from the QR-code 500 in each odd-numbered row, and inserting a blank fully transparent pixel to the left of each pixel from the QR-code 500 in each even-numbered row of the code 500.
  • Figure 6 is a flowchart: of an example process 600 for analyzing a possibly encoded source image to predict whether the image likely does or does not encode a watermarking image, in some implementations, the process 600 is performed by the image and analysis and decoder module 118 in cooperation with watermark predictive model 132 and controller 06 of the server system 102.
  • the system obtains a possibly encoded source image in some implementations, the possibly encoded source image is acquired by a user capturing a screenshot of content displayed in a webpage or graphical interface of an application at the user's client computer device.
  • the web page or application that is the target of the screens hot may include a source image that is susceptible to being encoded with a vistially-imperceptile watermarking image, such as the source image 128a from third-party content provider 106 being encoded with a watermarking : image 126 that was derived from QR-code 500.
  • the user may manually (or the user's device may automatically) capture the screenshot and submit if for analysis by the server system lor any number of reasons, such as for an analytics program, to report offensive or irrelevant content in the source image, to redeem promotional content encoded in the image, or to otherwise identify information about the source or context of the source image being served to the client computer 104.
  • the source image may or may not be merged into a single image file when rendered at the client computer 104, but typically after the user has captured a screenshot of the source image the effect is to create a merged image of any partially transparent watermarking image laid over top of the source image.
  • one or more factors related to the circumstances in which the screenshot was captured may increase the difficulty in the server's ask of decoding the image.
  • the user's device may have rendered a zoomed ⁇ in or zoomed-out view of the image at the time the screenshot was Captured such that the image is displayed at a higher or lower resolution than the original source image and watermarking images, and the server system may receive no indication o f the zoom level and corresponding resoluti on of the image.
  • the screenshot may over- or trader-capture the possibly encoded source image.
  • the possibly encoded source image is a banner ad in a webpage for which the user submits a screenshot of the entire webpage
  • the screenshot may be cropped and contain less than all of the possibly encoded source i mage.
  • the screenshot may include noise as a result of file compression that reduces the storage and/or transmission expense of the screenshot
  • the system generates one or more sealed versions of the possibly encoded source image.
  • the system may he configured to perform decoding of a possibly encoded source image at its native, original resolution (e.g., a 100-percent zoom level such that the size of the possibly encoded source image matches the original size of the source image) to ensure predictabili ty'- in the pattern of encoded and non-encoded pixels.
  • the system generates a set of scaled versions of the possibly encoded source image, where each scaled image provides a di fferent version of the possibly encoded source image at an estimated resolution that corresponds to the 100-perceni zoom level (also referred to as the “target’ * resolution or zoom level) by assuming a different starting zoom level for the possibly encoded image that was obtained af stage 602 and reducing or expanding the size of the image from the ass umed starting zoom level to the target 100-perceM zoom level.
  • the 100-perceni zoom level also referred to as the “target’ * resolution or zoom level
  • the system would generate a first scaled version of the possibly encoded source image by reducing the dimensions of the image in half to achieve the target 100-perceni zoom level. If the screenshot were assumed to have been captured at a 500- percent zoom level, the system would generate a second scaled version of the possibly encoded source image by reducing the dimensions of the image by one-fifth to achieve the target 100-percent zoom level. Any number of scaled versions of the possibly encoded source image may be generated.
  • the system Because many user devices from which a screenshot of a possibly encoded source image may be captured provide the capability for a user to set zoomed-in display resolutions at any resolution from 100-percent to 500-percent zoom levels, in some implementations the system generates respective scaled versions of the possibly obtained source image corresponding to all 401 possible starting zoom levels from 100- to 500-percent However, in other implementations, a different range of zoom levels may be assumed as possible starting resolutions. Additionally , in some implementations, only a subset of possible zoom levels within a gi ven range are sampled and used as a starting resolu tion from which a scaled version of the possibly encoded source image is generated.
  • the system selects a first scaled version of the possibly encoded source image for a first estimated starling zoom level, and searches for possibly encoded regions in the scaled image. If the starting zoom level is estimated to the target zoom level (e.g., 100-percent), then the selected image may simply be the original image that was obtained at stage 602.
  • a possibly encoded region of the selected/scaled image is a region that is determined to possibly he encoded with one or more instances of the watermarking image.
  • watermarking images are much smaller than the source image itself and the watermarking images are therefore : tiled directly adjacent to each other without overlap over all or a portion of the source image.
  • the watermarking images must be continuous with each other (e.g., a series of directly adjacent/neighboring watermarks) . If two instances of a possibly encoded watermarking image are detected in the scaled image that are non-contiguous with each other (either directly or indirectly via other instances of possibly encoded watermarking images), then they may he assigned to distinct possibly encoded regions.
  • Figure 9 illustrates an example of a set of possihly encoded regions 902a-902d within a scaled version 900 of a possibly encoded source image 950. The dashed lines indicate divisions between blocks corresponding to different instances of a possibly encoded watermark image within a particular region,
  • stage 702 the system (e.g , image analysis and decoder module 118) obtained the scaled version of the possibly encoded source image that is to be searched for possibly encoded regions.
  • stage 704 the system predicts encoded values of the possibly encoded pixels in the scaled image.
  • the system may predict the encoded values of every other pixel in the scaled image, i.e., pixels that would correspond to an encoded pixel in any watermarking image(s) that scaled image encoded.
  • the encoded value of each encoded pixel may be predieted by analyzing the color of that pixel relati ve to the colors of its neighboring blank pixels.
  • the pixel may be predicted to encode a black pixel from the watermarking image (e.g., a value of '1') If the color of the possibly encoded pixel is not sufficiently darker than one or more of its neighboring pixels or an average of its neighboring pixels, then the pixel may he predicted to encode a white pixel from the watermarking image (e.g., a value of 'O’), in some implementations., rafter than applying a binary classification to each possibly encoded pixel to predict whether the pixel encodes either a black pixel from the watermarking or encoding image or a white pixel from fte watermarking or encoding image, the system may generate a score (e.g., a probability value) for each pixel that indicates a likelihood that the
  • stage 706 the system determines possibly encoded regions of the scaled image based on patterns of the predicted encoded values of the possibly encoded pixels in the sealed image.
  • the system identifies a possibly encoded region by evaluating the predicted encoded values of the possibly encoded pixels in the scaled image to identify distinctive patterns indicative of a watermarking or encoding image. When the predicted encoded values match the predefined pattern for a watermarking or encoding image, an instance of a watermarking or encoding image is registered as a block for a possible encoded region.
  • the system may search for paterns of predicted encoded values that correspond to the distinctive square patterns 502a-502c of fte QR-code 500, or may search for a portion of the square patterns 502a-502c such as the pattern black- white-black-black-black-white-Mack (e.g , values 1-0-1- 1-1 -0-1) from the middle three rows of each square pattern. If these pre-defined patterns are recognized in the predicted encoded values, then the system registers an instance of a possibly encoded watermarking or encoding image in the scaled image at the location where fte patterns were recognized.
  • the pattern black- white-black-black-black-black-white-Mack e.g , values 1-0-1- 1-1 -0-1
  • the system may require perfect correspondence between the predicted encoded values and pre-defined patterns to register a match in other implementations, the system may permit some degree of tolerance so that imperfect correspondence between the predicted encoded values and the pre-defined patterns can nonetheless result in a registered match (e.g., so long as at least a threshold portion of the predicted encoded values match the corresponding encoded values pro vided in the pre-defined paterns). Contiguous sets of one or more identified blocks of possibly encoded watermarking/encoding images in the scaled image ate then assigned to respective possibly encoded regions.
  • the process of searching for and iden tifying possibly encoded regions can be enhanced through performance of the operations of stages 708-712 of Figure 7.
  • the system determines a sequence score for each (or a subset) of the possibly encoded pixels in the scaled image.
  • the sequence score for a given possibly encoded pixel indicates a likelihood that the encoded values for a sequence of possibly encoded pixels that includes or has a defined relationship to the given possibly encoded pixel matches a predetermined distinctive pattern of encoded values that Is indicative of the presence of a watermarking or encoding image
  • the sequence score for a given possibly encoded pixel Indicates a likelihood that the encoded values for foe next seven possibly encoded pixels in the same row of the scaled image as the given possibly encoded pixel matches the black-whlte- black-black-black-white-black (e.g , 1-0-1-1-1-0-1 )sequence that is found In the middle three rows of the "square patterns 502a-502c of QR-code 500
  • the sequencescore is determined based on a numberof matches of the predicted encoded values to the predetermined sequence of encoded values.
  • the sequence score is determined based on consideration of the individual probability values for each possibly encoded pixel in the sequence relative
  • the system identifies every possible posi tion of a block, i.e., a possibly encoded instance of a watermarking or encoding image, in the scaled image, and for each block position, a block score indicating a likelihood that a watermarking or encoding image is encoded at that block position.
  • the number of block positions is approxi mately equal to the number of pixels in the scaled image since a di fferent block position cm be identified by sliding the block pixel-by-pixel (except perhaps around the periphery of the image where a lull block may not be possible as it would extend beyond the edges of the image).
  • the block score for a block position can be computed based on the sequence scores for certain posi tions of possibly encoded pixels in the block. For example, to determine a block score for a block at a given position of the scaled image that indicates a likelihood that a watermarking image corresponding to the QR-code 500 has been encoded in the block at that position, the system may evaluate the sequence scores for pixels near the top-left, top-right, and bottom-left portions ofthe block that indicate a likelihood that the seven pixel sequence from the square patterns 502a-502c occur at the appropriate positions of the block that would correspond to an encoded QR-code watermarking image.
  • the block score may indicate a higher likelihood that a watermarking image is encoded in the block, for example, if the sequence scores at those three comers of the block each indicate a high likelihood that the pixels at those comers have been encoded with the black-white-black- black- black-white- black (e.g., 1 -0-1 -1-1 -0-1) sequence
  • the system determines possibly encoded regions of the scaled image based on the block scores. For example, for each block position in the scaled image, the system may assess the likelihood that one or more contiguous blocks starting from that block position are encoded with respecti ve instances of a watermarking or encoding image. From the block scores and analysis of the likelihoods of various permutations of contiguous blocks each being encoded with a watermark, the system resolves a set of mutually exclusive possibly encoded regions for which the system has sufficient confidence that the regions are likely encoded with one or more contiguous bl ocks of watermarking or encoding images.
  • the system det ermines values for a set of predictive features for the sealed image. These values will be subsequently processed to generate an ultimate or final prediction as to -whether the possibly encoded source image has been encoded with a visually-imperceptible, partially transparent watermarking image (or set of repeated watermarking images tiled across the source image) according to the techniques described previously with respect to Figures 2-4
  • the predictive features characterize information about the possibly encoded regions identified in the scaled image.
  • the features may include (i) a mean (or other average) number of blocks in the identified possibly encoded regions of the scaled image, (ii) the number of blocks in the largest identified possibly encoded region of the scaled image (i.e., a maximum number of blocks among the possibly encoded regions), (in) a standard deviation or other measure of spread in the number of blocks contained in the identified possibly encoded regions of the scaled image, (iv) a total number of identified possibly encoded regions in the seated image, or a combination of two or more of those,
  • values for additional features are computed based on analysis of pixels in the unsealed version of the possibly encoded source image.
  • the system may map the identified possibly encoded regions from the scaled version of the possibly encoded source image to corresponding regions in the unsealed version of the possibly encoded source image.
  • One such mapping is illustrated by way of example in Figure 9, where each of the possibly encoded regions 902a-902d from the scaled version 900 of the possibly encoded source image is mapped to a corresponding region 952a- 952d in the unsealed version 950 of the possibly encoded source image.
  • the sy stem For each mapped region 9S2a-952d in the unsealed version of the possibly encoded source image, the sy stem computes three values: (i) variance in the red channel values for the pixels in the mapped region, (ii) variance in the blue channel values for the pixels in the mapped region, and (iii) variance in the green channel values for the pixels in the mapped region. The system then selects the lowest (minimum) blue channel variance among the mapped regions 952a- 952d, the lowest green channel variance among the mapped regions 952a-952d, and the lowest blue channel variance among the mapped regions 952a ⁇ 952d.
  • Var[S] and Var[N] are non-negative, and Var[X] > 0.
  • Var[Z] when Var[Z] is small, it is largely due to VarjXj.
  • Var[Z] when Var[Z] is small, it can be appreciated that Var[X] can be approximated by using Var[Z] in the case where watermarks are present.
  • Var[X] is small, greater confidence can he achieved in the classificationresult (e.g. / whether watermarks are encoded in the source image).
  • Var[X] is large, less confidence may be seen in the classification result.
  • the system stores the values of the predictive features that were determined at stage 608.
  • the system maintains a database of values of predict ive features determined for each scaled version of the possibly encoded source image that the system analy zes Each time new values for a set of features is determined for another scaled version of the image, a new record or entry may he created in the database and the values stored therein.
  • An example database table 1000 for storing predictive feature values associates with a given possibly encoded source image is depicted in Figure 10 As shown each row in the table 1000 represents a record that stores values of features for adifferent scaled version of the possibly encoded source image.
  • the system may temporarily store values in the database for the features computed for each scaled version of the image until feature values have been computed for all zoom levels of interest.
  • the system determines whether feature val ues are to he compu ted for any addi tional scaled images corresponding to differen t estimates of the starting zoom level for the possibly encoded source image. If additional scaled images are to be analyzed:, the system proceeds to stage 611, selects a next sealed version of the possibly encoded source image, and returns to stage 606 If no additional scaled images remain to be analyzed, the system proceeds to stage 614.
  • the system accesses the database of feature values computed for each of the scaled images and selects one of the sets of feature values for a particular scaled image as a representative set.
  • Various rules can be applied in determining which set of feature values to select as the representative set.
  • the system selects the set of feature values for the scaled version of the possibly encoded source imago that yielded the largest possibly encoded region in terms of the number of blocks in the region.
  • the set of feature values corresponding to this scaled version of the Image may be selected as the representative set. Additional or other features may also be used as a basis for selecting the representative set.
  • multiple sets of feature values for several zoom ratios may be selected as representative sets such as the top n number of sets (e.g... n - 2, 3, 4, or 5), In some implementations, all of the sets of feature values may he selected as representative sets.
  • the system processes the values in the representative set(s) of predicti ve features with a predictive model, e.g., watermark predictive model 132, to generate a final prediction indicati ve of whether the possibly encoded source image likely has or has not been encoded with a watermarking image.
  • a predictive model e.g., watermark predictive model 132
  • the predictive model is configured to generate abinary output indicating th at the possibly en coded source image either was likely encoded with a watermarking image or was likely not encoded with a watermarking image.
  • the predictive model is configured to output a score (e.g,, a probability value) indicating a numerical likelihood that the possibly encoded source image was encoded with a watermarking image or was not encoded with a: watermarking image in some implementations, the predictive model may generate the score as an intermediate output and then compare the score against a specified threshold to produce a final binary output indicating wh ether the possibly encoded source image likely was or was not encoded with a watermarking image. Any suitable form of predictive model may be employed for this task. In some implementations, the predictive model is a machine-learning model such as a neural network.
  • the neural network may be a feedforward neural network having an input layer configured to process the values from the representative set(s) of predictive features, one or more hidden layers that perform a series of operations on the inputs, and. an output layer that generates the predictive output.
  • An Illustration 800 of a watermark predictive model 802 processing values for seven features 864a-804g from a single representative set of values to generate a watermark prediction 806 is depicted in Figure 8. [0093 ]
  • the system can use the watermark prediction to inform decisions about whether to incur further processing expense in continuing to attempt to decode the possibly encoded source image.
  • a system controller determines whether to continue decoding based on the output of the predictive model if the prediction indicates that the possibly encoded source image likely was encoded with a watermarking image, then the system proceeds to stage 620 and uses another decoding model to attempt decoding of the possibly encoded source image. If successful, the encoded source image can be decoded to reconstruct the watermarking image, the encoding image (e.g., QR-code 500), or both. From the watermarking or encoding image, the plaintext data item and any information that it references can then be identified.
  • the encoding image e.g., QR-code 500
  • the system when decoding does proceed at stage 620, the system may use information gleaned from earlier stages of the process 600 to target specific regions for decoding that are estimated to have foe highest likelihood of success.
  • the targeted regions may he the identified possibly encoded regions corresponding foe scaled version of foe image whose feature values were sel ected as the representative set, for example
  • FIG 11 is a flowchart of an example process 1100 for training a predictive model, such as a neural network that implements the watermark predictive model 132 employed in foe server system 102.
  • foe process 1100 is performed by the model training engine 134.
  • the system obtains a set of training images. Some images in the training set have been encoded with a watermarking image while others have not, bin each image is labeled with a target (true) classification value indicating whether the image has or has not been watermarked.
  • the system determines values for predictive features for each training image (stage 1104).
  • the predictive features may include, for example, the seven features described with: respect to Figures 6 and 8.
  • the system then creates a set of training samples (stage 1106).
  • Each training sample corresponds to adifferent one of the training images and contains a training input and a target training output.
  • the training input includes or consists of values for foe predictive features determined for foe corresponding training image.
  • the training output is a value representative of the target output of the predic tive model given foe training input.
  • the target output can be a binary value that Indicates whether or not the corresponding training image from which the values for the predictive features were derived was or was not encoded with a watermarking image.
  • the system initializes internal parameters (e.g , using a randomized initialization process) of the neural network model (stage 1108).
  • the system selects a first training sample (stage 111:0) and then processes the training input from that training sample according to the current values of the model 's internal parameters to generate a predictive output (stage 3112).
  • the system determines any error between the predictive output and the target output of the training sample (stage 1114), and then uses the error to update the internal parameters of the model so as to reduce the error (stage 1116) ⁇ in some implementations, the error can he back-propagated and the model updated using gradient descent.
  • the system checks whether additional training samples remain to be processed or if training has completed if the trainig phase is not yet complete and additional samples remain, the process 1100 returns to stage 1110 where a next training sample is selected and operations 1112-118 are repeated.
  • the predictive model can be provided for use, e.g., as watermark model 132 or 802 in practice (stage 1120).
  • Implementations of the subject matter and the functional operations described in this specification can be implemented in digital electronic circuitry, in tangibly-embodied computer software or firmware, in computer hardware, including the straetnres disclosed in this specification and their structural equi valents, or in combinations of one or more of them, implementations of the subject matter described in this specification can be implemented as one or more computer programs, re., one or more modules of computer program instructions encoded on a tangible non-transitory program carrier for execution by, or to control the operation of, data processing apparatus.
  • the program instructions can be encoded on anartificially-generated propagated signal, e.g , a machine-generated electrical, optical, or electromagnetic signal, which is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus.
  • the computer storage medium can be a machine-readable storage device* a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them.
  • data processing apparatus refers to data processing hardware and encompasses all kinds of apparatus, devices, and machines for processing data, including by way of example a programmable processor, a computer, or multiple processors or computers.
  • the apparatus can also be or further include special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit).
  • the apparatus can optionally include, in addition to hardware, code that creates an execution environment for computer programs, e.g., code that constitutes processor firmware, a protocol stack., a database management system, an operating system, or a combination of one or more of them.
  • omputer program which may also be referred to or described as a program, software, a software application, a module, a software module, a script, or code
  • omputer program can be written in any form of programming language, including compiled or interpreted languages, or declarati ve or procedural languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.
  • a computer program may, but need not, correspond to a file in a file system.
  • a program can be stored in a porton of a file that holds other programs or data, e.g., one or more scripts stored in a markup language document, in a single file dedicated to the program in question, or in multiple coordinated files, e.g., files that store one or more modules, sub-programs, or portions of code.
  • a computer program can be deployed to he executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communication network.
  • Computers suitable for the execution of a computer program incl ude by way of example, general or special purpose microprocessors or both, or any other kind of central processing unit.
  • a central processing unit will receive instructions and data from read-only memory or a random access memory or both.
  • a computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices for storing data, e.g., magnetic, magneto-optical disks, or optical disks.
  • mass storage devices for storing data, e.g., magnetic, magneto-optical disks, or optical disks.
  • a computer need hot have such devices.
  • a computer can be embedded in another device, e.g,, a mobile telephone, a: personal digital assistant (PDA), a mobile audio or video player, a game console, a Global Positioning System (GPS) receiver, or a portable storage device, e.g., a universal serial bus (USB) flash drive, to name just a lew,
  • PDA personal digital assistant
  • GPS Global Positioning System
  • USB universal serial bus
  • Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks.
  • semiconductor memory devices e.g., EPROM, EEPROM, and flash memory devices
  • magnetic disks e.g., internal hard disks or removable disks
  • magneto-optical disks e.g., CD-ROM and DVD-ROM disks.
  • the processor and the memory can he supplemented by, or incorporated in, special purpose logic circuitry.
  • implementations of the subject matter described to this specification can be implemented one computer having a display device, e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor.
  • a display device e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor.
  • a keyboard and a pointing device e.g , a moose or a trackball, by which the riser can provide input to the computer.
  • Oilier kinds of devices can be used to provide for interaction wi th a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g.hav visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input.
  • a computer can interact with a user by sending documents to and receiving documents from a device that is used by the user; for example, by sending web pages to a web browser on a user's device in response
  • Implementations of foe subject matter described in this specification can be implemented in a computing system that includes a back-end component, e g., as a data .server, or that includes a middleware component, e,g., an application server, or that includes a front-end component, e.g , a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the subject matter described to this specification, or any combination of one or more such back-end, middleware, or front-end components.
  • the components of the system can be interconnected by any form or medium of digital data communication, e.g , a communication , network. Examples of communication networks include a local area network (LAN) and a wide area network (WAN), e,g., the Internet.
  • LAN local area network
  • WAN wide area network
  • the computing system can include clients and servers.
  • a client and server are generally remote from each other and typically interact through a communication network.
  • the relationship of client antiserver arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other in some implementations, a server transmits data, e.g , an HTML page, to a user device, e.g , for purposes of displaying data to and receiving user input from a user interacting with the user device , which acts as a client
  • Data generated a t the user device e,g., a result of the user interaction, can be received from the user device at the server,
  • FIG. 12 shows a schematic diagram of a generic computer system 1200.
  • the system can be used for the operations described in association with any of the computer-implemented methods described previously, according to one implementation.
  • the system 1200 includes a processor 1210, a memory 1220, a storage device 1230, and an inpui/output de vice 1240.
  • Each of the components 1210, 1220, 1230, and 1240 are interconnected using a system bus 1250
  • the processor 1210 is capable of processing instructions for execution within the system 1200.
  • the processor 1210 is a single-threaded processor.
  • the processor 1210 is a multi-threaded processor.
  • the processor 1210 is capable of processing instructions stored in the memory 1220 or on the storage device 1230 to display graphical information for a user interface on the input/output device 1240.
  • the memory 1220 stores information within the system 1200.
  • the memory 1220 is a computer-readable medium.
  • the memory 1220 is a volatile memory unit.
  • the memory 1220 is a non-volatile memory unit.
  • the storage device 1230 is capable of providing mass storage for the system 1200, in one implementation, the storage device 1230 is a computer-readable medium, la various different implementations, the storage device 1230 may he a floppy disk device, a hard disk device, an optical disk device, or a tape device.
  • the input/output device 1240 provides input/output operations for the system 1200.
  • the input/output device 1240 includes a keyboard and/or pointing device.
  • the input/output device 1240 includes a display unit for displaying graphical user interfaces.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Image Processing (AREA)
  • Editing Of Facsimile Originals (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Image Analysis (AREA)

Abstract

L'invention concerne des systèmes, des procédés, des dispositifs et d'autres techniques pour détecter, à l'aide d'un modèle prédictif, la présence d'une image filigrane dans une image source capturée. Les techniques peuvent impliquer des opérations qui consistent à obtenir une première image, la première image étant une version mise à l'échelle de l'image source capturée. Des couleurs de pixels dans la première image sont analysées pour déterminer des valeurs codées de pixels éventuellement codés dans la première image, et des motifs de valeurs codées des pixels éventuellement codés dans la première image sont analysés pour déterminer les régions éventuellement codées de la première image. Des valeurs pour des caractéristiques des régions éventuellement codées de la première image sont déterminées, et un modèle prédictif traite les valeurs de caractéristiques pour générer une prédiction indiquant si une image filigrane est codée dans la première image. La prédiction est stockée et fournie à un dispositif de commande pour prendre une décision concernant l'appel à un processus de décodage.
EP19773283.7A 2019-09-06 2019-09-06 Détection de filigranes d'image semi-transparents Pending EP3815034A1 (fr)

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/US2019/050011 WO2021045781A1 (fr) 2019-09-06 2019-09-06 Détection de filigranes d'image semi-transparents

Publications (1)

Publication Number Publication Date
EP3815034A1 true EP3815034A1 (fr) 2021-05-05

Family

ID=68000123

Family Applications (1)

Application Number Title Priority Date Filing Date
EP19773283.7A Pending EP3815034A1 (fr) 2019-09-06 2019-09-06 Détection de filigranes d'image semi-transparents

Country Status (6)

Country Link
US (1) US20210334929A1 (fr)
EP (1) EP3815034A1 (fr)
JP (1) JP7142709B2 (fr)
KR (1) KR102423710B1 (fr)
CN (1) CN112789650B (fr)
WO (1) WO2021045781A1 (fr)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230325961A1 (en) * 2021-06-21 2023-10-12 Google Llc Zoom agnostic watermark extraction
EP4172925A1 (fr) * 2021-06-21 2023-05-03 Google LLC Extraction de filigrane agnostique de zoom
CN116745799A (zh) * 2022-01-11 2023-09-12 谷歌有限责任公司 端到端水印系统
CN114440834B (zh) * 2022-01-27 2023-05-02 中国人民解放军战略支援部队信息工程大学 一种非编码标志的物方与像方匹配方法
US11706369B1 (en) * 2022-03-02 2023-07-18 Avalara, Inc. Systems and methods for digitally watermarking resources produced by an online software platform
CN114863143B (zh) * 2022-05-20 2023-03-21 北京百度网讯科技有限公司 图像检测方法、装置、电子设备及存储介质
US20240037688A1 (en) * 2022-07-27 2024-02-01 GeoSnapshot Pty Ltd Inspection of digital images through watermarks
CN116385414B (zh) * 2023-04-10 2023-11-07 哈尔滨耐是智能科技有限公司 零部件多尺度缺陷特征提取方法、计算设备及存储介质

Family Cites Families (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5832119C1 (en) * 1993-11-18 2002-03-05 Digimarc Corp Methods for controlling systems using control signals embedded in empirical data
US6360000B1 (en) * 1998-11-09 2002-03-19 David C. Collier Method and apparatus for watermark detection for specific scales and arbitrary shifts
US7013021B2 (en) * 1999-03-19 2006-03-14 Digimarc Corporation Watermark detection utilizing regions with higher probability of success
JP2002092683A (ja) * 2000-07-14 2002-03-29 Oki Electric Ind Co Ltd 媒体真偽鑑別装置
US6952485B1 (en) * 2000-09-11 2005-10-04 Digimarc Corporation Watermark encoding and decoding in imaging devices and imaging device interfaces
MXPA05003984A (es) * 2002-10-15 2005-06-22 Digimarc Corp Documento de identificacion y metodos relacionados.
JP4269861B2 (ja) * 2003-09-12 2009-05-27 沖電気工業株式会社 印刷物処理システム，透かし入り文書印刷装置，透かし入り文書読み取り装置，印刷物処理方法，情報読み取り装置，および情報読み取り方法
WO2012034280A1 (fr) * 2010-09-16 2012-03-22 Hewlett-Packard Development Company, L.P. Tatouage numérique
CN103748900B (zh) * 2011-02-04 2018-03-09 斯诺弗雷克解决方案公司 用于数字媒体的独特水印的方法及系统
JP2013009162A (ja) * 2011-06-24 2013-01-10 Mitsubishi Electric Corp 画像電子透かし埋め込み装置及び画像電子透かし検出装置
US9715659B2 (en) * 2011-08-26 2017-07-25 Google Inc. System and method for determining a level of confidence that a media item is being presented
JP6216516B2 (ja) * 2013-02-25 2017-10-18 株式会社日立ソリューションズ 電子透かし埋め込み方法および電子透かし検出方法
JP6138706B2 (ja) * 2014-01-17 2017-05-31 株式会社日立ソリューションズ 電子透かし埋め込み方法または電子透かし検出方法
US9805434B2 (en) * 2014-08-20 2017-10-31 Verance Corporation Content management based on dither-like watermark embedding
US9756401B2 (en) * 2015-04-07 2017-09-05 Sony Corporation Processing and providing an image in which a plurality of symbols are encoded
CN106991423B (zh) * 2016-01-21 2019-06-07 腾讯科技（深圳）有限公司 识别规格图片中是否包含水印的方法及装置
WO2017130334A1 (fr) * 2016-01-27 2017-08-03 楽天株式会社 Dispositif de traitement d'images, procédé de traitement d'images, et programme
US20170329942A1 (en) * 2016-05-12 2017-11-16 Markany Inc. Method and apparatus of drm systems for protecting enterprise confidentiality
IL251149A0 (en) 2017-03-14 2017-06-29 Google Inc Semi-transparent marks for transferred content

Also Published As

Publication number Publication date
KR20210030247A (ko) 2021-03-17
JP7142709B2 (ja) 2022-09-27
WO2021045781A1 (fr) 2021-03-11
CN112789650B (zh) 2024-03-12
KR102423710B1 (ko) 2022-07-22
CN112789650A (zh) 2021-05-11
JP2022503307A (ja) 2022-01-12
US20210334929A1 (en) 2021-10-28

Similar Documents

Publication Publication Date Title
EP3815034A1 (fr) Détection de filigranes d'image semi-transparents
Zhou et al. Coverless image steganography using partial-duplicate image retrieval
Ni et al. ESIM: Edge similarity for screen content image quality assessment
Walia et al. Digital image forgery detection: a systematic scrutiny
US10867171B1 (en) Systems and methods for machine learning based content extraction from document images
US9317676B2 (en) Image-based CAPTCHA exploiting context in object recognition
Zhang et al. Viscode: Embedding information in visualization images using encoder-decoder network
US9177199B2 (en) Semantic magazine pages
CN108288064B (zh) 用于生成图片的方法和装置
US8478035B2 (en) Method and apparatus for engaging functionality with a color tag
US20230362399A1 (en) End-to-end watermarking system
CN115240203A (zh) 业务数据处理方法、装置、设备及存储介质
US20230328101A1 (en) Systems and methods of detecting anomalous websites
US20230325961A1 (en) Zoom agnostic watermark extraction
US20230325959A1 (en) Zoom agnostic watermark extraction
CN113821677A (zh) 一种生成封面图像的方法、装置、设备及存储介质
US20240073478A1 (en) Determining video provenance utilizing deep learning
Michaylov Exploring the Use of Steganography and Steganalysis in Forensic Investigations for Analysing Digital Evidence
US20240087075A1 (en) End-to-end watermarking system
Yan et al. MSA-Net: Multi-scale attention network for image splicing localization
CN117496525A (zh) 图像中文字的质检方法、装置、电子设备及可读存储介质
CN112085670A (zh) 一种联合语义可解释信息的图像修复方法及系统

Legal Events

Date Code Title Description
STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: UNKNOWN

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE

PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE

17P Request for examination filed

Effective date: 20200715

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: EXAMINATION IS IN PROGRESS

17Q First examination report despatched

Effective date: 20221109

DAV Request for validation of the european patent (deleted)
DAX Request for extension of the european patent (deleted)